Subject selection Testing procedure

3 Intelligibility

3.1 Recorded Text Testing RTT

We did RTT in each dialect area to ascertain mutual intelligibility between the speakers of different dialects.

3.1.1 Test development and administration

First, short stories about an interesting event in the story-teller’s life were recorded in the vernacular in five dialects, namely Toff, Ambul, Richa, Kamwai, and Marhai-Massenge. The stories were about two to three minutes in length. These texts were transcribed, and then an English phrase-by-phrase free translation was made with key words glossed. For each of these texts, 15 basic comprehension questions were created. No inferences or background knowledge were required to answer the questions correctly. The correct responses were found in the texts themselves. The questions attempted to cover a wide range of semantic categories e.g. who, what, when, where, how many, action, instrument, quotation, etc. and were formed in such a way as not to be answerable on the basis of common knowledge alone. The texts were recorded on minidiscs, so track marks were inserted indicating points where comprehension questions should be asked. Standard RTT methodology Casad 1974 prescribes that the questions then be recorded in the speech variety of the subjects and inserted into the story following the information being asked about. This would ensure consistency of test administration and eliminate the variable of the subject’s command of a language of wider communication LWC from the test results. However, experience has shown that subjects tend to find responding to questions coming from a machine extremely unnatural and rather confusing. Responding to a person asking questions is much more natural and eliminates this confusion from the testing situation. Hence we verbally asked questions and did not insert the questions into the recording. We administered the RTT by playing the text up to the point where a question should be posed, pausing the playback, and then asking the subject the question. We asked the questions in English or Hausa, sometimes followed by a translation into the local dialect, provided by local interpreters. The subject responded either in the local dialect, Hausa, or English. The response was written down in English, Hausa, or Kulere. When appropriate, the tester would prompt to clarify if the subject understood that portion of the text. For example, if the subject responded with other elements of the story but not the answer sought by the question, the question was posed again or rephrased slightly. The tester, as well as the interpreter, had to take care not to give away the correct answer in their prompt questioning and interpreting. If there was some distraction during the playback of a portion of the text, such as loud noises, people talking, or the subject had a problem understanding the text the first time, the tester replayed that portion of the text. All such follow-up questions and replays were noted on the answer sheets. Since the tester is working with a rough transcription of a text in an unfamiliar speech variety, it is possible that the questions devised will not coincide with the text or appear to be logical to native speakers. The initial fifteen questions were pilot tested with ten speakers of the same speech variety as the story. Questions which elicited unexpected, wrong, andor inconsistent responses from vernacular speakers were eliminated, and the ten best questions were retained for the final test.

3.1.2 Subject selection

For the pilot test we found at least 10 subjects of any age most were young and gender who were vernacular speakers of that variety and were willing to participate. When administering the actual test, we tested people with as little contact with the other varieties as we could find. The guidelines for acceptable levels of inherent intelligibility are based on testing people with no contact with the other varieties. This is necessary to determine the inherent intelligibility of the test dialect as opposed to intelligibility acquired through contact. Pupils in primary school class 3 P3 to junior secondary school classes 3 JSS3 were chosen as test subjects. Where possible, we tried to choose subjects in P4–P6 or junior secondary school classes 1–3 JSS1–JSS3. The students in this age range are old enough to have a good command of their dialect but have not usually had many opportunities to visit other villages. Therefore, they are likely to be those with the least amount of contact with other speech varieties. We selected a sample of 10 vernacular speakers of each speech variety but not from the villages where the story is taken to listen to the texts and take the test.

3.1.3 Testing procedure

The testing procedure for the RTT on this survey was in four parts: subject identification and screening, a practice or familiarization test, a “hometown” test, and intelligibility testing. The first phase consisted of gathering basic demographic information age, sex, level of education, verifying that the subject was a vernacular speaker of the dialect of interest mother and father were both from the dialect area and screening for contact with the other dialects being tested subject does not travel to one of the other dialect areas frequently or have close relatives from another dialect. With only a few exceptions due to a shortage of subjects all subjects met these criteria. The second phase was a short practice test in the subject’s own dialect, which served to familiarize the subject with the testing procedure. The practice story was written out by a surveyor and then translated and recorded in order to save time on transcription and to make sure the story contained useful content. Four questions were created for this story and the test was not graded as part of the RTT. This was sometimes played twice if it appeared the subject was beginning to understand the procedure, but had not completely understood. This was done to screen out poor test takers. The third phase was the hometown test. Each subject was tested using the text in his own dialect first, which served as a further screening procedure. If the subject could not score 85 on the test in his own dialect, he was judged incompetent to follow the testing procedure or incompetent in his vernacular dialect and no further testing was performed. The final phase was administering the test in the dialects of interest. After passing the hometown test, each subject was tested on the three other stories. The order in which the other tests were played was rotated to avoid any effect on the average scores by the subject’s test-fatigue.

3.1.4 Scoring